1. Field of the Invention
The present invention relates to thin film deposition, and more particularly to a method and apparatus for monitoring the growth rate of single crystal films on stationary or revolving substrates, in which the thickness of the thin film is measured in terms of monolayer units.
2. Description of the State of the Art
Molecular beam epitaxy (MBE) is a versatile, thin film growth technique used in preparing high-quality, single crystal thin film materials and structures for fundamental studies and for manufacturing electronic devices, such as semiconductors. In order to engineer and obtain precise quality control over thin film components of microelectronic devices for desired band gaps, doping, junction structure, and the like, it is necessary that the thicknesses of these films be precisely monitored as they are grown to a predetermined number of atomic layers or monolayers. Such thicknesses are usually discussed in angstrom units (.ANG.), but the raw measurement can be made by monitoring the crystalline layers as they grow. Since it is known for each material, such as silicon, how thick each crystalline layer or monolayer is in angstrom units, monitoring and identifying each new crystal lattice layer as it is grown can provide the basis for calculating the overall thickness in angstrom units. Also, by determining the number of layers grown per unit of time and multiplying by the angstroms of thickness per monolayer can yield angstroms of growth per unit of time. Thus, the device can be designed by time of deposition with a particular constituent flux to yield the desired crystalline thickness in the film deposited. Consequently, accuracy within the range of one or two monolayers is needed.
There have been significant research efforts expended on developing real-time methods of determining and monitoring when a monolayer has in fact been deposited, and to thereby control the supply of the vapor phases of the materials being deposited.
Of the various methods developed for measuring the thickness of a crystalline film that is being grown on a substrate in a molecular beam epitaxy (MBE) chamber, a reflection high energy electron diffraction (RHEED) system is the most desirable. In a conventional RHEED system, a beam of high-energy (5 to 40 KeV) electrons is directed at the surface of the crystalline film growing in the MBE chamber at a grazing angle (one to two degrees to the plane of the surface). Because of the shallow incident angle, the electrons penetrate the surface of the growing crystalline film only a few atomic layers, and the resulting diffraction pattern of the electron beam is projected onto a phosphor screen positioned at the other side of the substrate. The projected pattern of electrons impinging the phosphor screen is therefore indicative of the crystalline surface, giving a measure of crystalline quality as well as a number of lattice conditions. It has further been established that once crystalline growth is initiated, there are oscillations in electron impingement intensity on portions of the phosphor screen related to fractional changes in surface atomic coverage. These temporal oscillations correspond to the growth of crystalline layers of the film and, when plotted versus time, can show a growth rate of the thin film.
In appropriate conditions, the deposition of materials, such as GaAs or other III-V compounds, proceeds in a crystal growth model characterized layer by layer as the incoming flux condenses on the crystalline substrate, so the resulting RHEED output oscillations can be correlated to development or growth of individual monolayers of the film. This current model of crystal growth explains the RHEED pattern oscillation as well as inevitable dampening of the oscillations as the film grows thicker. Prior to deposition the equilibrium surface is assumed to be smooth, which equates with minimal surface roughness, thus maximum reflectivity. When growth is initiated, submonolayer coverage produces island-like clusters randomly deposited over the surface of the substrate forming a somewhat "rough" appearance. Consequently, where the starting smooth surface caused little scattering of the electron beam and a light, intense pattern of electron incidence on a spot or location on the phosphor screen, the island-like crystals reduce reflectivity and tend to scatter the reflected electron beam. When the nascent crystalline layer is half-filled with atoms, the surface is at its "roughest," thus the reflectivity is at a minimum and electron beam scattering is at a maximum. The result is a decrease in intensity of electron incidence on that original projected spot or location on the phosphor screen. As more atoms are deposited and the monolayer nears completion, the roughness of the surface decreases again, and the resulting scattering of the incident electron beam also decreases. Therefore, the intensity of electron impingement on the original projected spot or location on the phosphor screen increases again. Consequently, a plot of intensity at the original projected spot on the phosphor screen versus time shows one oscillation of electron impingement intensity between two adjacent high values for each monolayer deposited. However, the surface of the newly deposited layer does not become entirely smooth, because the next succeeding layer often nucleates and starts to grow prior to the complete formation and lattice fill of the currently forming layer. Therefore, the intensity differential, detected on the phosphor screen, between the maximum and minimum in reflectivity gradually decreases after each successive layer, thus giving rise to a damping effect of each successive intensity oscillation. This cycle repeats itself over and over, giving rise to RHEED oscillations and the gradual damping of the same. Consequently, while one period of the intensity oscillation corresponds to exactly one monolayer, each successive oscillation becomes less pronounced and more difficult to detect or read. The result is that conventional RHEED oscillations are easy to detect and read for the first few crystalline layer formations; however, they rapidly become unreadable and useless for monitoring layer formation or crystalline growth.
There have been some apparatus and methods developed for monitoring the growth of thin films that do utilize RHEED oscillations, as described above. For example, G. W. Turner, et al. in a technical publication, entitled "Frequency-Domain Analysis of Time-Dependent Reflection High-Energy Electron Diffraction Intensity Data," J. Vac. Sci. Technol. B, Vol. 8, No. 2 pp 283-287, 1990, discloses a RHEED video system in which a video camera is focused on the phosphor screen. The video signal resulting from light produced by electron impingement on the phosphor screen is directed through a computer interface to a video monitor, where the image of the light emanating from the phosphor screen is displayed and broken down into a matrix of pixels. The intensity of these pixels, which are directly related to the intensity of the light emanating from the phosphor screen, are then stored as a function of time. The Eckstein et al., patent, U.S. Pat. No. 4,812,650 discloses a method of monitoring oscillations obtained by measuring the current density of photoemitted electrons in-situ, produced by irradiating the crystalline film with a hydrogen/deuterium UV light source. U.S. Pat. No. 4,855,013, issued to Ohta et al., discloses an improved method of monitoring approximately 400 oscillations by way of a RHEED measurement system, thus, allowing the growth to be measured for the first 1000 .ANG. of an AlGaAs alloy grown at an average growth rate of 5000 .ANG./hour.
The above technical paper and patents each have certain disadvantages, in regard to the combination of monitoring and growth techniques employed, as a trade off for certain advantageous features. For example, the video system utilized by Turner et al. has the advantage of being capable of monitoring the intensity of multiple light spots on the phosphor screen, thus electron diffraction patterns incident on the phosphor screen. However, it has a poor dynamic range, poor sensitivity, and poor signal to noise ratio, which results in a decrease in the number of oscillations, thus numbers of crystalline layers, capable of being monitored. Ohta et al. discloses a great improvement over the prior an in that the ability to monitor crystal growth was increased approximately ten fold. However, a relatively slow growth rate was employed (which enhances the number of observable oscillations). In addition, Al containing alloys (such as AlGaAs) sustain oscillations for a longer period of time as compared to GaAs. Further, to minimize uneven crystal growth, the substrate should be rotated during the thin film deposition, but both Turner et al. and Ohta et al. require stationary substrates and cannot accommodate rotating substrates. Eckstein et al. did rotate the sample substrate for greater uniformity; however, they were not able to use a RHEED measurement system because of the rotation. Consequently, Eckstein et al. could only monitor the growth rate to approximately 28 .ANG..
There is still a need, therefore, for a system or technique for monitoring crystal growth by layers very accurately for larger thicknesses during higher growth rates than has been achieved by the prior art. There is also a need for a system that can make such measurements at such accuracies in such high growth rates, but which can also accommodate rotating substrates.